International Journal of Pattern Recognition and Artificial Intelligence

The welding defect classification method based on deep learning often faces problems such as insufficient training data and complex model structures, which affect its real-time performance. Therefore, a welding defect classification method based on lightweight convolutional neural network (CNN) is proposed. The problems of insufficient and unbalanced welding defect image are solved by using the generative adversarial networks (GANs) data augmentation method. A lightweight CNN model is developed, which reduces the structural parameters under the premise of ensuring classification accuracy. The feature data of each convolution layer are visualized to verify the feasibility of the model and improve the interpretability of the model. By comparing the accuracy and real-time performance with other lightweight models, the excellent performance of the proposed model in welding defect classification is verified. Additionally, our model achieves 98.25% accuracy on the MNIST dataset.

To address the issues in clothing pattern style migration, this paper proposes a digital simulation and re-editing method for clothing patterns based on deep learning and somatosensory interaction. First, the proposed method encodes the black-and-white line drawing image, generating random noise images through a diffusion process, introducing color information for synthesis, and using a decoder to reconstruct a colored image. Afterwards, an improved VGG19 model is used to reconstruct content features and perform linear color transformation on style images, enabling pattern style migration through the construction of a Gram matrix and resulting in colored clothing texture patterns. Finally, a KinectV2 is utilized for fabric simulation, overlaying colorful clothing texture patterns to achieve 3D virtual dressing. The experimental results show that the proposed method improves the structural similarity index measure (SSIM) by 9–11% and the peak signal-to-noise ratio (PSNR) by 3–8% when compared to existing algorithms. The experiments provide evidence that the proposed method effectively mitigates color overflow, delivers precise image coloring, and accomplishes realistic restoration of clothing texture. Furthermore, the method offers an improved garment fit to fulfill the user’s interaction requirements.

As technological advancements progress and energy conservation and emission reduction policies gain traction, an increasing amount of clean energy is being integrated into the power grid system. This influx of new energy imposes stringent demands on the transmission lines within the power grid system. In recent years, the State Grid has implemented a plethora of intelligent transmission line inspection strategies, with the intelligent inspection of Unmanned Aerial Vehicle (UAV) transmission lines receiving significant promotion and widespread application. However, practical application has revealed that the prevalent transmission line detection algorithms yield a substantial quantity of false detections, particularly in the detection of nut defects in small-sized metallic fittings, voltage balancing ring defects, and defects in uninsulated conductors. To address these issues, this paper employs deep learning algorithms for target detection, critical point detection, and instance segmentation, focusing on aspects such as algorithmic logic, algorithmic models, and data processing. The aim is to enhance the precision of these three types of defect detection, diminish the rate of false detections, and augment the practicality of intelligent grid inspection.

Land cover segmentation has been a significant research area because of its multiple applications including the infrastructure development, forestry, agriculture, urban planning, and climate change research. In this paper, we propose a novel segmentation method, called Frequency-guided Position-based Attention Network (FPA-Net), for land cover image segmentation. Our method is based on encoder–decoder improved U-Net architecture with position-based attention mechanism and frequency-guided component. The position-based attention block is used to capture the spatial dependency among different feature maps and obtain the relationship among relevant patterns across the image. The frequency-guided component provides additional support with high-frequency features. Our model is simple and efficient in terms of time and space complexities. Experimental results on the Deep Globe, GID-15, and Land Cover AI datasets show that the proposed FPA-Net can achieve the best performance in both quantitative and qualitative measures as compared against other existing approaches.

We propose a full-reference image quality assessment (FR-IQA) method by incorporating low-level and high-level image features. First, in contrast to the preexisting deep IQA methods, which only use the features extracted by the deep network, we not only use the image gradient to replace the low-level features in the first two stages of the deep network, but also combine them with the middle-stage features of the deep network to construct the new low-level features. The deep features of shallow layers contain some unwanted noise and further result in a decline in IQA performance. Second, we combine the global features extracted by the self-attention-based model with the semantic features extracted by the convolutional neural network to form the high-level features. Instead of directly using the self-attention-based model trained on the classification task, we first train a no-reference (NR) IQA regression model on a larger dataset and then use the global features of this NR-IQA model. The self-attention-based model can capture the internal connections of the image and is more effective in extracting global information due to its larger perceptual field. In the final pooling stage, we combine the average pooling and the standard deviation pooling to obtain the dispersion and concentration of the similarity maps for a more comprehensive description of quality. Experiments show that our FR-IQA method is able to obtain competitive results on three standard IQA datasets.

Automated object measurement is becoming increasingly important due to its ability to reduce manual costs, increase production efficiency, and minimize errors in various fields. In this paper, we present a novel approach to three-dimensional (3D) object measurement based on point cloud modeling. Our method introduces a fast point cloud modeling computation framework consisting of five stages: coordinate centralization, rotation and translation, noise filtering, plane projection, and geometric computation. Furthermore, we propose a fast convex hull optimization algorithm to reduce the high complexity problem of traditional convex hull calculation. Our extensive experiments demonstrate that our approach outperforms existing methods in terms of measurement error rate and time savings, with a maximum time saving of 31.03% under certain error conditions.

Image segmentation is an essential part of medical image processing, which plays a significant role in adjunctive therapy, disease diagnosis, and medical assessment. To solve the problem of insufficient extracting context information, especially for medical image segmentation, this paper proposes a novel network architecture of multi-scale object context dilated transformer network (Multi-OCDTNet) to improve the utilization and segmentation accuracy for context information. The multi-scale object context transformer module can extract the multi-scale context information of the image through a three-layer transformer structure in a parallel way. The dilated convolution self-aware module can enhance the awareness of multi-scale context information in the feature map through layering transformer block groups and a set of transformer layers. In addition, we propose a composite weight-assigned-based loss function based on DDCLoss and Focal Tversky Loss to improve the stability of the segmentation performance of Multi-OCDTNet by adjusting the weight. The performance of Multi-OCDTNet is validated on the DRIVE and STARE datasets with segmentation accuracy of 97.17% and 97.84%, respectively, indicating the Multi-OCDTNet network possesses a significant competitive advantage in improving the segmentation performance of retinal vessel images.

Rapid and accurate confirmation of operational intent can improve the efficiency of flight operators in handling sudden emergencies. A lightweight DenseNet flight operator blink interaction confirmation method based on visual gaze is proposed to address the issues of low accuracy and long-time consumption in traditional methods for confirming operational intentions. After determining whether the visual gaze of the flight operator is located at the confirmation button, this method uses two consecutive blinks to confirm the intention of the operation. When performing blink recognition, an improved lightweight DenseNet network model is used, which only uses one block and introduces attention SEBlocks in its $L$1 and $L$2 layers. Through testing experiments on our own dataset with open and closed eye images, it has been shown that this method has higher accuracy and lower loss compared to the classical DenseNet for open and closed eye recognition. The interactive experiment of flight operation intention confirmation showed that this method requires a shorter time for operation intention confirmation.

One of the biggest difficulties, which China’s power infrastructure is now experiencing, is the capacity to estimate demand of short-term electricity. It can accurately estimate changes in the total power load in particular locations. Accurate forecasting results, which also serve as a dependable guide for power system operation, may increase the flexibility and resource usage of the contemporary power market. Power load characteristics are impacted by several things. This study presents an estimate method for short-term demand of electricity based on WOA-BILSTM in order to comprehensively evaluate the time series features presented in the power load data and improve the accuracy of power load forecasting. After extracting features for the variables that influence power load, this method uses the bidirectional long-term and short-term memory (BILSTM) neural network layer for bidirectional time series feature learning. Using the local electricity load data from Quanzhou from 2018 as the data set, the prediction model is constructed by screening the multi-dimensional input parameters and selecting the feature vectors with good association as the input carefully. By comparing the outcomes of three popular load forecasting models, including LSTM network, BILSTM network, and WOA-BILSTM, it is demonstrated that the WOA-BILSTM neural network method is more accurate and effective than others two. By building an optimal combination model, this technique could increase the accuracy of power load data’s short-term predictions. At the same time, it also reduces the time of personnel debugging.